Based On Multi-source Data The Modeling Of Regional Forest Productivity

Forest is the largest carbon stock of terrestrial ecosystem and forest productivity is an significant parameter of forest carbon sequestration.The quantification of forest productivity has been the focus of many scientific studies?e.g.carbon cycle,climate change,etc.?.The traditional estimation method of forest productivity was based on the data of field investigation.And the ring data has the advantage on accurate dating,long time series and widely distributed samples.Since the arrangement of sample plot was affected by topography,climate and other reasons,the estimation method of forest productivity was limited on the regional scale.With the development of remote sensing technology,remote sensing data had some advantages on macro monitoring and was applied to estimate GPP of terrestrial ecosystem over the regional scale for a relatively long period.The MOD₁7 model was one of the most popular models but the model parameters had some limitations.The simpler model,termed the Temperature and Greenness?TG?model,was based on enhanced vegetation index?EVI?and land surface temperature?LST?and could simulate forest productivity better than the MOD-7 model.In this paper,the study areas were located in Qilian Mountains and Northeast China.And the method of dendrochronology,MOD₁7 model and TG model were implemented to estimate forest productivity.The main conclusions were summarized as follows:?1?Savizky-Golay?S-G?time series filtering method.Because of sensor,cloud,aerosol,snow and other factors,the original MODIS EVI and LSI data would appear noise or missing data.The S-G filtering method could effectively remove image noise and available describe the detail features of the troughs and peaks.And the filtering method could reconstruct the missing data of LST but the LST data may be lifted after S-G filtering.?2?Estimation of forest aboveground net primary productivity?ANPP?from tree ring data and vegetation index.After S-G filtering to the EVI time series data,the accuracy of the regression model?iEVI-ANPP?was improved available.This study indicated that the method that combining the tree ring data and EVI to estimate Picea crassifolia AINPP was practically feasible.From 2000 to 2013 forest ANPP increased more slowly,with an average value of 135.47 g m-2 and the average annual amount of 1052.57 t.Furthermore,the result also revealed a distinct spatial pattern in variability whereby forest ANPP increased from the northwest to the southeastern.Forest ANPP was positively correlated with precipitation?R=0.49?and temperature?R=0.22?.Compared with temperature,the precipitation was the material cause to affect the dynamic change of forest ANPP.?3?MOD₁7 model test and improvement at Changbaishan?CBS?site.The validation of CBS flux data showed that the MOD 17 model underestimated the tower GPP in growth season.It could be possible to correct problems with the current version of MOD₁7 model by enhancing the precision of the meteorological and other data inputs.However,to a certain extent the uncertainty parameter of MOD₁7 model limited its own development.?4?The parameter calibration of TG model and the inversion of forest GPP in Northeast China.The TG model results were much closer to the tower GPP values than were the MODIS GPP product results.And the inversion precision of evergreen forest was higher than deciduous forest.From 2000 to 2014 the annual average GPP of four forest types had a slight fluctuation in Northeast China.The annual forest GPP for deciduous broadleaf forest,mixed forest,deciduous needleleaf forest and evergreen needleleaf forest were 1737.71 g m-2,1458.53 g m-2,1248.83 g m-2 and 831.84 g m-2,respectively.The forest GPP generally increased from northwest to southeast and mainly distributed between 1000 g m-2 and 2000 g m-2.The annual average total GPP for mixed forest,deciduous broadleaf forest,deciduous needleleaf forest and evergreen needleleaf forest were 352.87 Tg,118.71 Tg,24.71 Tg and 0.20 Tg,respectively.